Node Lifecycle WorkingGroup - Getting Started
summary
Working Group aims to standardize node and pod lifecycle with concrete proposals: a declarative maintenance API and an eviction request API that lets controllers intercept disruptions and report progress; prototypes target integration with graceful node shutdown, checkpoint/restore flows and existing tools like Node Problem Detector and DRA device taints. Use cases include fleet operations on 50,000 GPUs causing 100–1,000 daily reboots and in-place upgrades; goal is interoperable signals and deadlines so controllers can coordinate drains, migrations and autoscaling.
AI-written summary · about
chapters (17)
- 0:00Node Lifecycle WG intro
- 1:03Pod lifecycle observability
- 3:53Pod scheduling constraints
- 5:19Running pod states
- 7:26Disruption intents scaling
- 8:56Node shutdown issues
- 10:48Eviction API proposal
- 15:07Node lifecycle challenges
- 18:36Node lifecycle at scale
- 21:26Cloud node upgrades
- 23:55WG goals efforts
- 26:43How to get involved with the working group
- 27:27WG future vision
- 29:12Extending lifecycle beyond nodes
- 30:36Stuck finalizers cleanup
- 31:41Declarative node maintenance
- 33:26Node lifecycle abstractions
related talks
37mUnconference Session: Node Lifecycle State Needs a Real API
30mSupporting Long-Lived Pods Using a Simple Kubernetes Webhook
30mWG App Def - Intro
19mGraceful Node Shutdown – A Beginner’s Journey of Kubernetes Development
26mWriting a Custom Controller: Extending the Functionality of Your Cluster [I]
43mKubernetes SIG Node Intro and Deep Dive
28mIntro to the Kubernetes Working Group for Multi-tenancy
35mDeep Dive: Kubernetes Working Group for Multi-tenancy